A TeaSoftware product

Prototype

Hanmelo

Speak Mandarin with confidence.

Hanmelo is an AI-powered Mandarin speaking companion designed for Vietnamese learners. Our goal is to make speaking practice feel natural, accessible, and personal through voice conversations, useful feedback, and learning experiences that adapt over time.

Learners study Mandarin, but rarely get to speak it.

Vietnamese learners can build vocabulary, study grammar, and complete exercises on their own. What they usually lack is regular speaking practice in a low-pressure setting.

Speaking partners and tutors are not always available, and practicing in front of other people can feel intimidating. Without frequent conversation, confidence grows slowly even when knowledge grows fast.

A call, not a quiz.

  1. map

    Step 1

    Choose a scenario

    Ordering at a café, asking for directions, introducing yourself at work.

  2. call

    Step 2

    Start a call

    Talk with an AI conversation partner using your voice.

  3. record_voice_over

    Step 3

    Speak naturally

    Make mistakes and keep the conversation going. Corrections can wait until the end.

  4. fact_check

    Step 4

    Get feedback

    After the call, review mistakes, better phrasings, and what to practice next.

Post-call analysis with Claude.

We plan to use Claude, Anthropic's language model, to analyze each conversation after the call: identifying mistakes, extracting vocabulary and patterns worth reviewing, and personalizing future practice sessions.

This section describes the planned architecture. The Claude integration is under development and is not yet available in the prototype.

mic

Voice call

Learner speaks with the AI partner in a chosen scenario.

subject

Transcript

The conversation is converted to text.

psychology

Claude analysis

Mistakes, corrections, and review items are extracted.

auto_stories

Learning memory

Feedback shapes the next scenarios and review sessions.

Where Hanmelo is today.

  1. check_circle In planning
  2. radio_button_checked Prototype Current stage
  3. radio_button_unchecked Private Alpha
  4. radio_button_unchecked Public release

Next milestones

  • arrow_right Post-call transcript analysis through the Claude API.
  • arrow_right A set of test scenarios to evaluate feedback quality.
  • arrow_right A short recorded demo of a full practice session.
  • arrow_right A published data-handling policy before opening to external testers.

Voice data deserves care.

The Hanmelo prototype is not open to the public. Before we invite external testers, we will publish how voice recordings, transcripts, and learning memory are stored, used, and deleted.

Want to try Hanmelo early?

Email us to receive development updates and an invitation when testing opens.

mail Get updates